The Prediction Of The Possibility Of Undergraduates' Internet Loans Based On User Portrait Technology

At present,the demand for online loans for college students is increasing.Online loans have become the object of consideration when college students need funds.However,due to lack of self-control and social experience,college students are often prone to over-consumption and over-use of online loans,resulting in failure to repay loans,resulting in tragedies.Society,schools,and parents often lag behind in accessing student online loan information.Therefore,the main purpose of this paper is to study the prediction of the possibility of college students' online loans.Through the relevant attributes of college students,build a predictive model,obtain information on the use of online loans by college students in advance,guide college students to understand financial management knowledge,reasonable online loans and inform their parents.To reduce the possibility of tragedies.This paper designs a questionnaire based on this purpose and collects information on online loans of students from three universities in Jiangxi Province.Firstly,study the characteristics of relevant attributes.Through descriptive statistical analysis and correlation analysis of the questionnaire,we find that 61.8% of college students have used online loan platform,and variables such as variable grade,monthly living expenses,and monthly living expenses balance use online loans.There is a significant relationship.Secondly,according to the user's portrait definition and neural network structure,the predictive model of college students' network loan possibility is constructed.Feature engineering is used to process features to obtain the required features of the model.Through the random grid parameter search algorithm and the ten-fold cross validation method,the optimal parameters of college students' online loan possibility prediction model are obtained.Finally,79% of the models were obtained,the accuracy rate was 77%,and the value was 79%.The results show that the model has a good forecasting effect on unknown data sets and can be used in practice to obtain information on college students' use of online loans in advance.